Apoorv Nayak Prathyusha Dasari Traffic Grooming. Agenda Improved approaches for cost effective traffic grooming in WDM ring networks Motivation

Apoorv Nayak

Prathyusha Dasari

Traffic GroomingTraffic Grooming

AgendaAgenda Improved approaches for cost effective traffic

grooming in WDM ring networks Motivation Terminology Single hop approach Multi hop approach

A Novel Generic Graph Model for Traffic Grooming in Heterogeneous WDM Mesh Networks

Motivation With WDM technology we can have dozens of

wavelengths on a fiber.

Increase in network capacity is accompanied with increase in the electronic multiplexing equipment.

Dominant cost is electronics and not fiber.

AimAim Goal is to minimize electronic costs by reducing

the number of ADM’s and make efficient use of wavelengths.

“Groom” a number of low rate traffic streams onto a higher rate stream and vice versa.

Reducing the number of wavelengths

TerminologyTerminology

SONET

ADM

WADM

SONET RingSONET Ring

Much of today’s physical layer infrastructure is built around SONET rings.

Constructed using fiber (one or two pairs usually used to provide protection) to connect SONET ADM’s.

ExampleExample

Signal from A split into two; one copy transmitted over the working ring (1) other copy over protection ring (8-7-6). B selects the best signal.

SONET ADMSONET ADM Add/Drop multiplexer.

Each ADM can multiplex multiple lower rate

streams to form a higher rate stream OR

demultiplex a higher rate stream to several

lower rate ones.

Employs O-E-O conversion.

Works at a particular wavelength.

ExampleExample

M<N

WADMWADM Wavelength add/drop multiplexer. Emergence of WDM technology has enabled a single

fiber pair to support multiple wavelengths. Since ADM works on a single wavelength, if there are W

wavelengths, every node would need N*W ADM’s.

WADM contdWADM contdBut a node may not need to add / drop streams on

every wavelength.

WADM’s can add/drop only the wavelengths carrying

traffic to/ from a node.

Example of a SONET ringExample of a SONET ring

OC-48 SONET ring

AssumptionsAssumptions

Traffic demands are static and known a priori.

Traffic is uniform;total bandwidth required is

same for any s-d pair.

Unidirectional ring considered.

Single hop approachSingle hop approach

Uses the simulated annealing heuristic.

A node with a wavelength-k ADM can communicate

directly with all other nodes having wavelength-k ADM.

Formation of a wavelength-k logical ring which consists

of the subset of N nodes with a wavelength-k ADM.

Nodes within a logical ring communicate with each other

directly (single hop).

1

52

34

Logical RingsLogical Rings

1

1

2

2 3

3

Example of single hop approachExample of single hop approachGiven data

Network layout Traffic demand matrix Number of available wavelengths : 2 Capacity of each wavelength : OC-3 Uniform traffic between any two nodes is OC-1.

Network TopologyNetwork Topology

0 1

3 2

0 1

23

fiber

t1

t2

t3

t4

t5 t6

a) Physical Network b) Traffic on the Network

Traffic Grooming Approach1 (Random)Traffic Grooming Approach1 (Random)

Total number of ADM’s needed = 8

Traffic Grooming Approach 2Traffic Grooming Approach 2

Total number of ADM’s needed = 7

Single hop traffic grooming algorithmSingle hop traffic grooming algorithm

do{do{

dcost = perturb();if(∆cost < 0 or (∆cost > 0 and exp(-∆cost/control) > rand

[0,1))) {

accept_change();chain++;

}else

reject_change();

} while(chain < ANN_CONST * G)

control = control * DEC_CONST;

} while(control > END)

TerminologyTerminology

Perturb() – Randomly swap positions of two circles in different

wavelengths.

ANN_CONST– Decides how long to run the algorithm before

system reaches equilibrium.

DEC_CONST - How fast to lower the control variable.

G – Grooming ratio (Ratio of the wavelength channel rate to the lowest traffic rate).

Multi hop approach (Hub based communication)Multi hop approach (Hub based communication)

Source and destination on different logical rings.

Solution OXC?

Still maturingCostly

Relatively inexpensive as compared to OXC More delay and reduced throughput Price-Performance tradeoff.

A “hub” node with an ADM for each wavelength. Multiple ADM’s at some nodes. Decide which nodes, how many ADM’s, which wavelengths.

Multiple ADM’s?

Approach followed in paper?

Example of a unidirectional multihop WDM ring networkExample of a unidirectional multihop WDM ring network

Terminology and assumptionsTerminology and assumptions W - Number of wavelengths.

Di - Number of ADM’s in the ith node.

G - Grooming ratio (Ratio of the wavelength channel

rate to the lowest traffic rate).

tij - Traffic requirement (Number of low rate circuits

between i and j for i-j pair).

tij = 1 for uniform traffic.

Given data

Number of nodes- NTraffic matrix- TGrooming ratio- G

ADM placement algorithmADM placement algorithm

Input N, G, t;Compute number of ADM’s needed at each node by the equation:

Compute number of wavelengths by the equation:

Create an ADM hub node;Place ADM’s needed at each node sequentially;

While (no of ADM’s and wavelengths can be reduced){

Assign traffic on each wavelength using shortest path;Traffic grooming (wavelength combining and segment swapping);

}

Wavelength CombiningWavelength CombiningIf capacity (i) < G and capacity (j) <GAnd capacity (i) + capacity (j) ≤ GThen the two wavelengths can be combined

2

3

4

W1 =1

W2 =2W1 =2

W2 =1

W1 =1

W2 =2

W1 =2

W2 =1

W1=3

W1=3 W1=3

W1=3

1

Segment SwappingSegment Swapping

Helps in wavelength combining by “manipulating” wavelengths such that all link capacities are less than G.

1

2

3

4

W1 = 2

W2 = 3

W3=1

W1 =3

W2 =2

W3=1

W1 =3

W2 =2

W3=1

W1 =2

W2 =3

W3=1

W1=3

W2=3

W1=3

W2=3 W1=3

W2=3

W1=3

W2=3

W1 =2

W2 =3

W3=1

W1 =2

W2 =3

W3=1

Example of multihop approachExample of multihop approach

Given data

N=5, node 0 is hub node

G = 3, tij = 1

By eqn 1, every node needs at least 2 ADMs

By eqn 2, total number of wavelengths is 8

0 1 2 3 4

3 3 3 3

3 3 3 3

3 3 3 3

3 3 3 3

1 1 1 1

1 1 1 1

1 1 1 1

1 1 1 1

3 3 3 3

2 2 2 2

Final result

Number of wavelengths = 4

Number of ADM’s = 12

ComparisonsComparisons

Increase in G, decrease in W, less ADM’s in hub node

A Novel generic ApproachA Novel generic Approach Objectives Generic Graph Model Auxiliary Graph

- Vertices

- Edges IGABAG Example Grooming Policies

- Comparison

In Heterogenenous NetworksIn Heterogenenous Networks

CSC 778 Fall 2007

Traffic Grooming :• Can be applied to static or dynamic traffic grooming problem.• Each node is characterized by various parameters - Optical switching/multiplexing capabilities- wavelength/waveband/fiber. - Electronic switching/multiplexing grooming capabilities. - Availability of wavelength conversion. - Number of transmitters/receivers.

CSC 778 Fall 2007

• Traffic grooming problem may have various objectives - Minimize cost (transmitters/receivers). - Minimize overall traffic load. - Minimize maximum traffic on any light path. - Minimize maximum wavelengths on any fiber.

ObjectivesObjectives

CSC 778 Fall 2007

• Construct auxiliary graph Add vertices and edges corresponding to network elements.

- Links- Wavelength converters- Electronic ports (transmitters/receivers)

Assign costs to links based on objective• Run shortest path algorithm

Generic Graph Model Generic Graph Model

CSC 778 Fall 2007

• Input and output vertex for each wavelength layer at each node• Input and output vertex for lightpath layer at each node• Input and output vertex for access layer at each node

Auxiliary Graph - VerticesAuxiliary Graph - Vertices

Auxiliary Graph - EdgesAuxiliary Graph - Edges

CSC 778 Fall 2007

• Wavelength Bypass Edges (WBE) - From each input to output port on a given wavelength layer. - Optical wavelength switching capability• Grooming Edges (GmE) - From input to output port on access layer if grooming is available. - Electronic switching capability.• Mux Edges (MuxE) - From output port on access layer to output port on lightpath layer.• Demux Edges (DmxE) - From input port on lightpath layer to output port on access layer.

Auxiliary Graph - EdgesAuxiliary Graph - Edges

CSC 778 Fall 2007

• Transmitter Edges (TxE) - From output port on access layer to output port on wavelength layer if transmitter is available .• Receiver Edges (RxE) - From input port on wavelength layer to input port on access layer if receiver is available.• Converter Edges (CvtE) - From input port on wavelength layer 1 to output port on wavelength layer 2 if optical wavelength conversion is available.

Auxiliary Graph - EdgesAuxiliary Graph - Edges

CSC 778 Fall 2007

• Wavelength-Link Edges (WLE) - From output port on wavelength layer l at node i to input port on wavelength layer l at node j if wavelength l is available on the physical link between i and j• Lightpath Edges (LPE) - From output port on the lightpath layer at node i to the input port of the lightpath layer at node j if there is a lightpath from node i to node j

..

CSC 778 Fall 2007

WBE

GrmE

MuxEDmxE

TxE

RxECvtE

WLE

Auxiliary Graph - Edges

Integrated Grooming Based on the Auxiliary Graph Integrated Grooming Based on the Auxiliary Graph (IGABAG)(IGABAG)

CSC 778 Fall 2007

• Traffic demand: T(s,d,g,m) - s : source, d : destination, g: granularity, m: amount of traffic in units of g• Step 1: Delete edges with capacity less than g.• Step 2: Find shortest path p from output port on the access layer of s to the input port on the access layer of d.• Step 3: If p contains wavelength-link edges, set up corresponding lightpaths.• Step 4: Route traffic demand along path p. If the capacity of lightpaths along p is less than m, route the maximum amount possible.• Step 5: Restore edges deleted in Step 1.• Step 6: Update graph G.

ExampleExample

CSC 778 Fall 2007

• Wavelength capacity: OC-48• Each node has 2 transmitters/receivers• Granularity: OC-12• Request 1: T(1, 0, OC-12, 2) -> Lightpath on 1 from N1 to N0• Request 2: T(2, 0, OC-12, 1)• Request 3: T(1, 0, OC-48, 1)

Example..Example..

CSC 778 Fall 2007

CSC 778 Fall 2007

0

1 2

Example…Example…

Example – Single-Hop GroomingExample – Single-Hop Grooming

CSC 778 Fall 2007

• Request 2: T(2, 0, OC-12, 1) - new lightpath on 2 from N2-N1-N0

• Request 3: T(1, 0, OC-48, 1)

CSC 778 Fall 2007

Example: single-hop groomingExample: single-hop grooming

CSC 778 Fall 2007

0

1 2

Example: single-hop groomingExample: single-hop grooming

Example – Multi-hop GroomingExample – Multi-hop Grooming

CSC 778 Fall 2007

• Request 2: T(2, 0, OC-12, 1) - new lightpath on 1 from N2-N1 - Existing lightpath on 1 from N1-N0• Request 3: T(1, 0, OC-48, 1)

CSC 778 Fall 2007

Example: multi-hop groomingExample: multi-hop grooming

CSC 778 Fall 2007

0

1 2

Example: multi-hop groomingExample: multi-hop grooming

Grooming OperationsGrooming Operations

CSC 778 Fall 2007

Add New Lightpath(s)

Single-hop or multi-hop grooming

Operation 1 No Single-hop

Operation 2 No Multi-hop

Operation 3 Yes Single-hop

Operation 4 Yes Multi-hop

Grooming PoliciesGrooming Policies

CSC 778 Fall 2007

• Minimize the Number of Traffic Hops (MinTH) - Attempt Operation 1 - Attempt Operation 3 - Between Operation 2 and 4, choose the one with fewest logical hops• Minimize the Number of Lightpaths (MinLP) - Attempt Operation 1 - Attempt Operation 2 - Attempt Operation 3 or 4• Minimize the Number of Wavelength-Links (MinWL) - Attempt Operation 1 - Attempt Operation 2 - Between Operation 3 and 4, choose the one with fewer wavelength links

Ordering of Requests for Static CaseOrdering of Requests for Static Case

CSC 778 Fall 2007

• Least Cost First (LCF) - Establish least-cost request first - Cost = (weight of shortest path for demand)/(amount of traffic)• Maximum Utilization First (MUF) - Select connection with maximum utilization first - Utilization = (amount of traffic)/(number of hops on physical topology)• Maximum Amount First (MAF) - Select connection with largest traffic demand first

Comparison of Policies – Non Blocking ModelComparison of Policies – Non Blocking Model

CSC 778 Fall 2007

Comparison of Policies – Blocking ModelComparison of Policies – Blocking Model

CSC 778 Fall 2007

AcknowledgmentsAcknowledgments• Improved Approaches for Cost-effective traffic grooming in WDM Ring Networks : Uniform-Traffic Case; Wonghong Cho, Jian Wang, Biswanath Mukherjee

• A Novel generic graph model for traffic grooming in heterogenous WDM mesh networks; Hongyue Zhu, Hui Zang, Biswanath Mukherjee

•Traffic grooming algorithms for reducing electronic multiplexing costs in WDM rings; Angela L. Chiu, Eytan H. Modiano

• An effective and comprehensive approach for traffic grooming and wavelength in SONET/WDM rings; Xijun Zhang, Chunming Qiao

•Improved approaches for cost-effective traffic grooming in WDM ring networks: Non-uniform traffic and bidirectional ring; Jian Wang, V. Rao Vemuri

•Connection Oriented Networks, Harry Perros

Questions??Questions??